When content such as program or movie content is recorded on a recording medium, the video data of the content are coded by a coding method such as the MPEG (Moving Picture Experts Group) method to create a video stream. The audio data of the content are coded by a method such as the AC-3 method to create an audio stream. In the MPEG-2 system stipulated in ISO/13818, the video stream and the audio stream are multiplexed into a TS (Transport Stream). The video data or audio data in the video stream or audio stream are broken up into 188-byte source packets, which are the minimum units of access. In the description below, video streams and audio streams will also be referred to simply as ‘streams’.
A video stream is made up of GOPs (Groups of Pictures), where a GOP is about 0.5 seconds in terms of video reproduction time. A GOP comprises I-pictures obtained by intra-frame coding, P-pictures obtained by inter-frame predictive coding in the forward direction, and B-pictures obtained by bidirectional predictive coding (in the description below, the term ‘picture’ will be used as a general term for I-pictures, P-pictures, and B-pictures).
An I-picture is placed at the beginning of a GOP. The I-picture at the beginning of a GOP is also treated as an access point: a position at which random access to the video stream is possible. The I-picture at the beginning of every GOP does not necessarily become an access point; if a plurality of GOPs constitute one access unit, for example, then the I-picture at the beginning of the first GOP among the plurality of GOPs is set as the access point.
In trick reproduction modes such as the fast-forward mode in which the video content is viewed by skipping from picture to picture, or when a function such as time search is used to start the reproduction of the content from an intermediate point in the content specified by a time, in general, first an I-picture is decoded and reproduced. To perform trick reproduction etc. at higher speeds, it is necessary to detect the positions of the I-pictures and their constituent source packets quickly. The reason why trick reproduction starts with the decoding of an I-picture is that until an I-picture is decoded, it is not possible to decode other pictures.
The I-pictures in a stream are conventionally detected with reference to an EP_Map in which the display time information (PTS: Presentation Time Stamp) and I-picture positional information (SPN: Source Packet Number) are stored. An EP_Map is provided for every GOP (e.g., Patent Document 1).
Information concerning the size of the I-pictures may be added to the above PTS and SPN information, these data may be assembled into a table and stored in the EP_Map, and the table stored in the EP_Map may be referred to in order to detect the position and size of the I-picture (e.g., Patent Document 2).    Patent Document 1: Japanese Patent Application Publication No. 2002-158971 (pp. 38-40, FIG. 138)    Patent Document 2: Japanese Patent Application Publication No. 2004-201034 (pp. 11-12, FIG. 5)